| Keywords |
Particulate matter; Machine learning; Inland area; Costal area; Contribution |
| Abstract |
The scope of this work was to characterize PM concentration in inland and coastal areas in Korea and to examine the relative contribution of the outdoor environment and time factor. The coastal area was designated as the case where there was a measuring station within 10 km of the coastline, and the other areas were defined as inland areas. PM10·PM2.5 concentrations and other environmental factors were all measured by the national measuring station in units of one hour in 2019. Through the prediction accuracy analysis by machine learning algorithm using these data, it was confirmed that the boosted decision tree had the highest accuracy in PM analysis, and the accuracy was lowered in PM10 and coastal areas due to factors that were not reflected in the feature such as sea salt. After that, by using variable weights derived from linear regression, we found the main causes of the PM concentration increase were PM precursor and the seasonal characteristics of winter and spring. And main causes of PM concentration decrease were dew point temperature, wind speed and direction. Also, the PM increase or decrease due to environmental variables was larger in the inland area than in the coastal area. |